Theories of Learning

PSY 565

Study tips in addition to the learning objectives for Connectionist Models of Knowledge

In a connectionist model a node does not represent knowledge. A node is like a real neuron.

The knowledge is represented as a pattern of activation.

Real neurons send activation by “firing” an action potential. The activation can be excitatory or inhibitory.

The all or none principle”: A real neuron has a threshold of activation. It has to receive some total amount of activation before it will fire. When it receives activation, it either fires or not, but it always fires at the same strength. The strength of the activation it sends is the same whether the threshold of activation was just barely exceeded, or exceed by a lot.

Real neurons receive activation from many other neurons at the same time. The amount of activation received is the sum of all of the excitatory activation, minus all the inhibitory activation.

The synapse: At the end of the axon there is a terminal bud. The terminal bud of a sending neuron releases chemical neurotransmitters into the gap between it and the dendrite of a receiving neuron. The region that includes the terminal bud, the gap, and the receiving dendrite is called the synapse. The gap itself is often called either the synaptic gap or synaptic cleft.

Neural networks consist of nodes that are like real neurons and links or connections that are like synapses.

Neural networks can perform many tasks that for a human would be mental tasks. They can categorize instances, recognize text and speech, and perform analogical reasoning.

The neural network approach is just in its infancy. It already has been shown to be a very powerful explanatory model. It is especially useful because it can explain how the brain could carry out complex computations using just neurons. It is promising because the models not only do things that humans do mentally, they also have characteristics like the characteristics of a human brain.